ankile/real01c-insert-marker-d1-baseline-uniform-r1-25k-nocf-eval-sobol20
收藏资源简介:
该数据集是使用LeRobot创建的机器人控制数据集,专门用于机器人学习和强化学习任务。数据集包含来自Franka机器人的20个episodes,总计5219帧数据,帧率为15fps,覆盖1个任务。数据以Parquet格式存储,总数据文件大小为100MB,视频文件大小为200MB。数据集结构包括观察状态(如笛卡尔位置、关节位置、速度、夹爪位置等)、动作(如笛卡尔速度、位置、关节速度、位置、夹爪控制)、奖励、完成标志、成功标志、策略ID等特征。此外,数据集还包含来自两个摄像头(标识为18650758_left和25916956_left)的视频数据,分辨率为480x640,3通道(RGB),编码为h264,无音频。这些数据适用于机器人控制算法训练、行为克隆、强化学习模型评估等应用。数据集中还包含元数据如时间戳、帧索引、episode索引等,以支持时间序列分析。
This dataset is a robotics control dataset created using LeRobot, specifically designed for robot learning and reinforcement learning tasks. It contains 20 episodes from a Franka robot, totaling 5219 frames at a frame rate of 15fps, covering 1 task. The data is stored in Parquet format, with a total data file size of 100MB and video file size of 200MB. The dataset structure includes observation states (such as Cartesian position, joint position, velocity, gripper position, etc.), actions (such as Cartesian velocity, position, joint velocity, position, gripper control), rewards, done flags, success flags, policy IDs, and other features. Additionally, the dataset includes video data from two cameras (labeled 18650758_left and 25916956_left) with a resolution of 480x640, 3 channels (RGB), encoded in h264, and without audio. This data is suitable for applications such as robot control algorithm training, behavior cloning, and reinforcement learning model evaluation. The dataset also contains metadata like timestamps, frame indices, and episode indices to support time-series analysis.




